[PATCH] D118857: [DWARF][codegen] Fix for Aranges when split inlining is present

David Blaikie via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Wed Feb 2 16:24:12 PST 2022


dblaikie added inline comments.


================
Comment at: llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp:2941-2942
   // Filter labels by section.
   for (const SymbolCU &SCU : ArangeLabels) {
+    if (SkeletonCUs.count(SCU.CU))
+      continue;
----------------
ayermolo wrote:
> dblaikie wrote:
> > Should we avoid putting these entries in `ArangeLabels` to begin with, rather than filtering them out here?
> I looked at where this is added, and I think it would involve passing a flag through multiple levels of API. I wasn't sure what would be more acceptable. Just cleaning it up here, or doing that. This is more self contained.
Could you roughly summarize the layers of API it'd have to go through? There might be some nice way to structure it.


================
Comment at: llvm/test/DebugInfo/X86/fission-inline-aranges.ll:11-31
+; struct foo {
+;   template<typename T>
+;   static void f2();
+;   static void f3(...);
+; };
+;
+; void f1();
----------------
ayermolo wrote:
> dblaikie wrote:
> > Does this need templates, using declarations, and if/scopes, etc? Or could it be simplified a bit? If it can't be, might be worth some explanatory text about why all these features are needed to tickle the bug?
> I re-purposed fission-inline.ll. Maybe just add extra checks there?
Be best to make something simpler if possible, I think.

What's the DWARF (or source) that's of interest/necessary to reproduce the issue?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D118857/new/

https://reviews.llvm.org/D118857



More information about the llvm-commits mailing list